Folds 15 tests/fix-*.test.cjs regression files (191 test() blocks) into their module's main suite, per the wave decomposition of #3315 (H3 of epic #3053). 187 blocks land in 8 existing suites (4 exact-duplicate cases dropped, documented inline); 4 blocks move via git mv into 2 new suite files with no prior coverage to merge into. Zero production behavior change. Also tightens two H1 (#3313) ratchets that the fold's own file-count reduction moved past their grace window, per the ratchets' documented dual failure mode (a stale/too-loose baseline fails exactly like a novel violation): - lint-test-file-count.allowlist.json: removes the stale "audit" entry (folding fix-2766 into tests/uat.test.cjs drops that module back to its 2-file cap). - lint-allow-test-rule-refs.ceiling.json: lowers maxFiles 314 -> 309, the real post-fold high-water mark (gsd-test's own repo-baseline test caught this — CI, not a human, found it). Two orthogonal review passes (Standards+Spec code-review, isolated security-review) found and this commit fixes two issues before push: a genuinely-distinct #2287 test case (file-absent vs. file-present- resolved) that a prior fold pass had wrongly dropped as a duplicate — restored verbatim into tests/uat.test.cjs; and a missing same-line allow-test-rule citation on the #2196 block in tests/debug-session-management.test.cjs, added for consistency with its sibling #2257 block. lint-removed-but-needed also caught two stale doc references to the now-folded-away fix-2285-claude-orchestration-wiring.test.cjs filename (docs/adr/1143-claude-orchestration-capability.md, gsd-core/references/execute-phase-response-language.md) — updated both to point at tests/claude-orchestration.test.cjs, its new home. Co-authored-by: sim <sim@local>
This commit is contained in:
Tom Boucher
@@ -412,3 +412,265 @@ describe('bug #2002: offer_next checks CONTEXT.md before suggesting next step',
|
||||
});
|
||||
});
|
||||
}
|
||||
|
||||
|
||||
// ────────────────────────────────────────────────────────────────────────
|
||||
// Folded from tests/fix-3177-execute-phase-dispatch-claim.test.cjs — test hygiene #3334 (H3)
|
||||
// ────────────────────────────────────────────────────────────────────────
|
||||
{
|
||||
const { describe: __foldDescribe } = require('node:test');
|
||||
__foldDescribe("folded:fix-3177-execute-phase-dispatch-claim (test hygiene #3334 H3)", () => {
|
||||
// allow-test-rule: runtime-contract-is-the-product #3177 — the workflow markdown is loaded
|
||||
// verbatim into the agent's context and the matrix/descriptor ARE the negotiated host
|
||||
// contract; asserting agreement between those documents is behavioral, not source-grep.
|
||||
|
||||
'use strict';
|
||||
|
||||
const { test, describe } = require('node:test');
|
||||
const assert = require('node:assert/strict');
|
||||
const fs = require('node:fs');
|
||||
const path = require('node:path');
|
||||
const fc = require('fast-check');
|
||||
|
||||
const ROOT = path.join(__dirname, '..');
|
||||
const WORKFLOW = path.join(ROOT, 'gsd-core', 'workflows', 'execute-phase.md');
|
||||
const MATRIX = path.join(ROOT, 'docs', 'reference', 'host-integration-capability-matrix.md');
|
||||
const DESCRIPTOR = path.join(ROOT, 'capabilities', 'claude', 'capability.json');
|
||||
|
||||
const workflowText = () => fs.readFileSync(WORKFLOW, 'utf8');
|
||||
|
||||
/**
|
||||
* Value of `| <field> | <value> | …` inside the `## <host>` section of the matrix.
|
||||
*
|
||||
* Anchored on a whole heading LINE, not a substring: `## claude` must never match
|
||||
* `## claude-local`, and the section must end at the next `## ` heading so a field
|
||||
* absent from this host can never be answered from the next host's table.
|
||||
*
|
||||
* @param {string} matrix - full matrix document text
|
||||
* @param {string} host - section name, e.g. `claude`
|
||||
* @param {string} field - row label, e.g. `dispatch.background`
|
||||
* @returns {string|null} trimmed cell value, or null when the section or row is absent
|
||||
*/
|
||||
function matrixField(matrix, host, field) {
|
||||
const lines = matrix.split('\n');
|
||||
const start = lines.findIndex((l) => l.trim() === `## ${host}`);
|
||||
if (start === -1) return null;
|
||||
let end = lines.length;
|
||||
for (let i = start + 1; i < lines.length; i += 1) {
|
||||
if (lines[i].startsWith('## ')) { end = i; break; }
|
||||
}
|
||||
const row = lines.slice(start + 1, end).find((l) => l.startsWith(`| ${field} |`));
|
||||
if (!row) return null;
|
||||
return row.split('|')[2].trim();
|
||||
}
|
||||
|
||||
describe('#3177: execute-phase.md states Claude Code dispatch truthfully', () => {
|
||||
test('execute-phase.md never claims Claude Code Agent() blocks or returns synchronously', () => {
|
||||
// Row 1 — the failing-first regression. Both stale sentences, by their own text.
|
||||
const text = workflowText();
|
||||
const stale = ['blocks until complete', 'returns synchronously'];
|
||||
const present = stale.filter((phrase) => text.includes(phrase));
|
||||
assert.deepEqual(
|
||||
present, [],
|
||||
'execute-phase.md still asserts synchronous Claude Code dispatch. Claude Code backgrounds '
|
||||
+ 'subagents by default (v2.1.198+); `run_in_background: false` is the opt-out.',
|
||||
);
|
||||
});
|
||||
|
||||
test('the Claude Code dispatch bullet states the background-by-default model', () => {
|
||||
// Row 2 — the corrected sentence must actually SAY the true thing, not merely
|
||||
// omit the false one. A deletion would pass row 1 while teaching nothing.
|
||||
const bullet = workflowText()
|
||||
.split('\n')
|
||||
.find((l) => l.startsWith('- **Claude Code:**'));
|
||||
assert.ok(bullet, 'the <runtime_compatibility> Claude Code bullet must exist');
|
||||
assert.match(
|
||||
bullet, /backgrounded by default/,
|
||||
'the bullet must state that dispatch is backgrounded by default',
|
||||
);
|
||||
assert.match(
|
||||
bullet, /verify completion/,
|
||||
'the bullet must point at completion verification. This workflow deliberately backgrounds '
|
||||
+ 'its executors (the multi-plan path prescribes run_in_background: true), so the blocking '
|
||||
+ 'opt-out is not the guidance here — confirming completion is.',
|
||||
);
|
||||
assert.ok(
|
||||
bullet.includes('Agent(subagent_type="gsd-executor"'),
|
||||
'the bullet must still carry the dispatch mechanism the rest of the file depends on',
|
||||
);
|
||||
});
|
||||
|
||||
test('the workflow prose and the capability matrix agree on claude dispatch.background', () => {
|
||||
// Row 3 — the parity guard. This is the assertion that outlives the wording:
|
||||
// flip the matrix to `false` without touching the prose (or vice versa) and this reds.
|
||||
const declared = matrixField(fs.readFileSync(MATRIX, 'utf8'), 'claude', 'dispatch.background');
|
||||
assert.equal(declared, 'true', 'matrix must document claude dispatch.background');
|
||||
|
||||
const bullet = workflowText()
|
||||
.split('\n')
|
||||
.find((l) => l.startsWith('- **Claude Code:**'));
|
||||
const proseSaysBackground = /backgrounded by default/.test(bullet ?? '');
|
||||
assert.equal(
|
||||
proseSaysBackground, declared === 'true',
|
||||
'execute-phase.md and the host-integration matrix disagree about whether claude backgrounds '
|
||||
+ 'its subagent dispatch. They describe the same runtime; exactly one of them is wrong (#3177).',
|
||||
);
|
||||
});
|
||||
|
||||
test('the claude descriptor and the matrix agree on dispatch.background', () => {
|
||||
// Row 4 — the other half of the divergence class. The matrix is generated from
|
||||
// the descriptor, so this pins the generator's output to its input.
|
||||
const descriptor = JSON.parse(fs.readFileSync(DESCRIPTOR, 'utf8'));
|
||||
const declared = matrixField(fs.readFileSync(MATRIX, 'utf8'), 'claude', 'dispatch.background');
|
||||
assert.equal(
|
||||
String(descriptor.runtime.hostIntegration.dispatch.background), declared,
|
||||
'capabilities/claude/capability.json and the rendered matrix disagree',
|
||||
);
|
||||
});
|
||||
|
||||
test('the Codex orchestrator rule is not swept by the Claude Code correction', () => {
|
||||
// Row 5 — negative space. Codex dispatch IS synchronous. A regex sweep for
|
||||
// "return its result" would introduce a NEW falsehood here; this catches that.
|
||||
const text = workflowText();
|
||||
const codexRules = text
|
||||
.split('\n')
|
||||
.filter((l) => l.includes('ORCHESTRATOR RULE — CODEX RUNTIME'));
|
||||
assert.ok(codexRules.length >= 2, 'both Codex orchestrator rules must survive');
|
||||
for (const rule of codexRules) {
|
||||
assert.ok(
|
||||
rule.includes('Wait for the subagent to return its result'),
|
||||
'Codex dispatch is genuinely synchronous — its wait rule must not be corrected away',
|
||||
);
|
||||
}
|
||||
});
|
||||
|
||||
test('the Copilot and multi-plan dispatch rules survive the correction', () => {
|
||||
// Row 6 — negative space. These were already true and are adjacent to the edit.
|
||||
const text = workflowText();
|
||||
assert.ok(
|
||||
text.includes('- **Copilot:** Subagent spawning does not reliably return completion signals.'),
|
||||
'the Copilot bullet must survive verbatim',
|
||||
);
|
||||
assert.ok(
|
||||
text.includes('one at a time with `run_in_background: true`'),
|
||||
'the multi-plan wave prescription must survive verbatim',
|
||||
);
|
||||
assert.ok(
|
||||
text.includes('If `Agent` IS available (top-level Claude'),
|
||||
'the spawn mandate is derived from TOOL AVAILABILITY, not from blocking — it must survive',
|
||||
);
|
||||
});
|
||||
});
|
||||
|
||||
describe('#3177: matrix section extraction is bounded by its heading', () => {
|
||||
const matrix = () => fs.readFileSync(MATRIX, 'utf8');
|
||||
|
||||
test('section extraction — first row of the section', () => {
|
||||
// limit-1: the row immediately after the `## claude` heading is INSIDE.
|
||||
assert.equal(matrixField(matrix(), 'claude', 'embeddingMode'), 'imperative');
|
||||
});
|
||||
|
||||
test('section extraction — last row before the next heading', () => {
|
||||
// limit: the final row of `## claude` is still INSIDE.
|
||||
assert.ok(matrixField(matrix(), 'claude', 'dispatch.isolation').startsWith('harness-worktree'));
|
||||
});
|
||||
|
||||
test('section extraction — a row in the next section never leaks in', () => {
|
||||
// limit+1: a field absent from claude must be null rather than silently
|
||||
// resolved from `## codex` below it, and two hosts with different values
|
||||
// must never resolve to the same cell.
|
||||
const doc = matrix();
|
||||
assert.equal(matrixField(doc, 'claude', 'dispatch.background'), 'true');
|
||||
assert.equal(matrixField(doc, 'claude', '__definitely_not_a_field__'), null);
|
||||
assert.notEqual(
|
||||
matrixField(doc, 'claude', 'effortSurface'),
|
||||
matrixField(doc, 'kilo', 'effortSurface'),
|
||||
'two hosts with different values must not resolve to the same cell',
|
||||
);
|
||||
});
|
||||
|
||||
test('fc property: field extraction never leaks across ## boundaries', () => {
|
||||
const hostArb = fc.stringMatching(/^[a-z][a-z0-9-]{0,12}$/);
|
||||
const valueArb = fc.stringMatching(/^[a-z0-9]{1,10}$/);
|
||||
fc.assert(
|
||||
fc.property(
|
||||
fc.uniqueArray(fc.tuple(hostArb, valueArb), {
|
||||
minLength: 2, maxLength: 6, selector: ([h]) => h,
|
||||
}),
|
||||
fc.nat(),
|
||||
(sections, pick) => {
|
||||
const doc = sections
|
||||
.map(([host, value]) => `## ${host}\n\n| axis | value |\n| f | ${value} |\n`)
|
||||
.join('\n');
|
||||
const [host, value] = sections[pick % sections.length];
|
||||
// Exactly the requested section's value, never a neighbor's.
|
||||
assert.equal(matrixField(doc, host, 'f'), value);
|
||||
// A longer name that merely EXTENDS a real heading resolves to nothing.
|
||||
// `_` is outside hostArb's alphabet, so this probe can NEVER collide with
|
||||
// another generated section — the `-local` form could, and did.
|
||||
assert.equal(matrixField(doc, `${host}_x`, 'f'), null);
|
||||
},
|
||||
),
|
||||
{ numRuns: 200 },
|
||||
);
|
||||
});
|
||||
});
|
||||
|
||||
describe('#3177: debug.md dispatches its session manager in the foreground', () => {
|
||||
const DEBUG_WF = path.join(ROOT, 'gsd-core', 'workflows', 'debug.md');
|
||||
const debugText = () => fs.readFileSync(DEBUG_WF, 'utf8');
|
||||
|
||||
/**
|
||||
* Every fenced `Agent( … )` block in debug.md that dispatches the session manager.
|
||||
*
|
||||
* Anchored on a line that is exactly `Agent(` so the PROSE mention of
|
||||
* `Agent(subagent_type="gsd-debug-session-manager", …)` inside the blockquote at
|
||||
* :206 is not mistaken for a dispatch. Positional selection (first match wins) was
|
||||
* the original bug here: it silently checked the continue path while the
|
||||
* new-session path went unexamined.
|
||||
*/
|
||||
function sessionManagerDispatches(text) {
|
||||
const blocks = [];
|
||||
for (const m of text.matchAll(/^Agent\($/gm)) {
|
||||
const close = text.indexOf('\n)', m.index);
|
||||
if (close === -1) continue;
|
||||
const block = text.slice(m.index, close);
|
||||
if (block.includes('subagent_type="gsd-debug-session-manager"')) blocks.push(block);
|
||||
}
|
||||
return blocks;
|
||||
}
|
||||
|
||||
test('every session-manager spawn carries the run_in_background: false opt-out', () => {
|
||||
// #2196 required this dispatch be foreground and blocking so the orchestrator
|
||||
// receives the session summary inline; debug.md still says "Wait for it; do not
|
||||
// background it" and "Display the compact summary returned by the session
|
||||
// manager". Claude Code backgrounds subagents by DEFAULT, so that intent only
|
||||
// holds if each call states the opt-out explicitly — prose alone silently
|
||||
// reinstated the exact lost-handoff failure #2196 was filed to fix.
|
||||
const blocks = sessionManagerDispatches(debugText());
|
||||
assert.equal(
|
||||
blocks.length, 2,
|
||||
'debug.md dispatches the session manager on BOTH the new-session and continue paths; '
|
||||
+ 'a change to that count means a dispatch was added or removed and must be re-checked.',
|
||||
);
|
||||
for (const block of blocks) {
|
||||
assert.match(
|
||||
block, /run_in_background\s*=\s*false/,
|
||||
'every gsd-debug-session-manager dispatch must pass run_in_background=false — without '
|
||||
+ 'it Claude Code backgrounds the spawn and the compact summary never returns (#2196).',
|
||||
);
|
||||
}
|
||||
});
|
||||
|
||||
test('debug.md does not assert the spawn is inherently foreground', () => {
|
||||
// The old premise ("is FOREGROUND and BLOCKING") was a property claim about the
|
||||
// host, not an instruction — and it was false for the same reason as #3177.
|
||||
assert.ok(
|
||||
!debugText().includes('is FOREGROUND and BLOCKING'),
|
||||
'debug.md must not claim the Agent() call is inherently foreground; it must name the '
|
||||
+ 'run_in_background: false opt-out that actually makes it so.',
|
||||
);
|
||||
});
|
||||
});
|
||||
});
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user